Table 2.

RAG-HPO enables LLMs to have greater precision and recall compared to traditional dictionary-based concept recognition approaches.

Analysis Tool	True Pos	False Pos	False Neg	Precision	Recall	F1	Comparison to RAG-HPO
							Precision	Recall	F1
RAG-HPO	1368	257	424	0.84	0.78	0.80	N/A	N/A	N/A
Doc2HPO	643	251	1150	0.70	0.36	0.46	****	****	****
ClinPhen	642	364	1150	0.63	0.36	0.43	****	****	****
FastHPOCR	820	721	973	0.53	0.47	0.48	****	****	****

RAG-HPO paired with Llama-3 70B correctly identified significantly more HPO terms compared to other assessed programs with a smaller effective false positive rate.

^****indicates p<0.00001.

Abbreviations: Neg, negative; Pos, positive.